Launch Plans

Tags: Basic

Launch plans bind a partial or complete list of inputs necessary to launch a workflow, along with optional run-time overrides such as notifications, schedules and more. Launch plan inputs must only assign inputs already defined in the reference workflow definition.

import calendar

When To Use Launch Plans

• For multiple schedules of a workflow with zero or more predefined inputs

• To run a specific workflow but with a different set of notifications

• To share a workflow with set inputs with another user, allowing the other user to simply kick off an execution

• To share a workflow with another user, making sure that some inputs can be overridden if needed

• To share a workflow with another user, ensuring that some inputs are not changed

Launch plans are the only means for invoking workflow executions. A ‘default’ launch plan will be created during the serialization (and registration process), which will optionally bind any default workflow inputs and any default runtime options specified in the project flytekit config (such as user role, etc).

The following example creates a default launch plan with no inputs during serialization.

import datetime

from flytekit import LaunchPlan, current_context, task, workflow


@task
def square(val: int) -> int:
    return val * val


@workflow
def my_wf(val: int) -> int:
    result = square(val=val)
    return result


default_lp = LaunchPlan.get_default_launch_plan(current_context(), my_wf)
square_3 = default_lp(val=3)

The following shows how to specify a user-defined launch plan that defaults the value of ‘val’ to 4.

my_lp = LaunchPlan.create("default_4_lp", my_wf, default_inputs={"val": 4})
square_4 = my_lp()
square_5 = my_lp(val=5)

It is possible to fix launch plan inputs, so that they can’t be overridden at execution call time.

my_fixed_lp = LaunchPlan.get_or_create(
    name="always_2_lp", workflow=my_wf, fixed_inputs={"val": 2}
)
square_2 = my_fixed_lp()
# error:
# square_1 = my_fixed_lp(val=1)

Putting It All Together

Default and fixed inputs can be used together to simplify individual executions and programmatic ones. Here is a simple example to greet each day of the upcoming week:

@task
def greet(day_of_week: str, number: int, am: bool) -> str:
    greeting = "Have a great " + day_of_week + " "
    greeting += "morning" if am else "evening"
    return greeting + "!" * number


@workflow
def go_greet(day_of_week: str, number: int, am: bool = False) -> str:
    return greet(day_of_week=day_of_week, number=number, am=am)


morning_greeting = LaunchPlan.create(
    "morning_greeting",
    go_greet,
    fixed_inputs={"am": True},
    default_inputs={"number": 1},
)

# Let's see if we can convincingly pass a Turing test!
today = datetime.datetime.today()
for n in range(7):
    day = today + datetime.timedelta(days=n)
    weekday = calendar.day_name[day.weekday()]
    if day.weekday() < 5:
        print(morning_greeting(day_of_week=weekday))
    else:
        # We're extra enthusiastic on weekends
        print(morning_greeting(number=3, day_of_week=weekday))